unsloth/phi-4
Same weights, repacked vs microsoft/phi-4. Same weights in different files (re-saved, re-sharded or new metadata), plus 200 new tensors.
Tensors changed: Embeddings 0/1, Norms 0/81, MLP 0/40, Attention 0/40, lm_head 0/1
16 derivatives checked
- mlx-community/Unsloth-Phi-4-4bit: Quantization
- shisa-ai/shisa-v2.1-unphi4-14b: Full fine-tune
- kholiavko/test-ph4: Full fine-tune
- kaisser/LLM-Maroc: Quantization
- dumbequation/Phi-4Garu-telugu-14b: LoRA-style
- shisa-ai/shisa-v2-unphi4-14b: Full fine-tune
- shisa-ai/ablation-34-rafathenev2.unphi45e6-shisa-v2-unphi-4-14b: Full fine-tune
- rAIfle/WAIDWML-Phi4-8x14B-bf16: MoE built from dense models
- shisa-ai/ablation-35-rafathenev2.unphi42e6-shisa-v2-unphi-4-14b: Full fine-tune
- ykarout/phi4-reasoning-merged: LoRA-style
- 1024m/HPhi-4-00-fp16: LoRA-style
- Yujivus/Phi-4-Health-CoT-1.1: LoRA-style
- shisa-ai/ablation-114-bestofn.atlo.rp.tlx.newmix2.unphi-shisa-v2-unphi-4-14b: Full fine-tune
- shisa-ai/ablation-87-bestofn.atlo.rp.tl.unphi42e6-shisa-v2-unphi-4-14b: Full fine-tune
- shisa-ai/ablation-179-finalsft2-shisa-v2-unphi-4-14b: Full fine-tune
- 1024m/PHI-4-B50: LoRA-style
5 other repos hold byte-for-byte identical weight files.